1. C. Cuvelier and P. Thunis JRC, European Commission Ispra - Italy Harmonisation in AQ Modelling Fairmode, Cavtat, 10 Oct 2008

2. Overview Harmonisation, support to AQ policy CityDelta project –Objectives, Methodology EuroDelta project –Objectives, Methodology

3. Need for Harmonisation Need for Evaluation of the uncertainty in AQ modelling used for AQ policy (see Directive AAQ). Is there a (consistent) difference between the fine scale (FS, 5km) and the large scale (LS, 50km) models? Do AQ models produce consistent responses to emission-reduction policy scenarios; What are the ranges of these responses ?  What is the robustness of AQ modelling results in the policy context?

4. Harmonisation as we see it Analysis of individual Modelling results (Validation against Obs) Intercomparison of Modelling results Ensemble approach (mean/median of the Models) Analysis of the variability around the Ensemble Validation of the Ensemble model Spin-off: improvement and further development of models The Ensemble model is a useful tool in the frame of the Harmonisation of the individual model results. It is often more robust and looks therefore more appropriate for policy purpose 

5. CityDelta

6. Policy context: Clean Air For Europe (CAFE) Launched in 2001 by the European Commission, CAFE was a programme of technical analysis aiming at the development of a long-term, integrated policy advice to protect against negative effects of air pollution on human health and the environment. Resulting in the Thematic Strategy published Sept 05 Question: Which measures will lead to a cost-effective reduction of air-pollution health-related problems in European Cities? In particular for O3 and PM CityDelta Objective: How to include sub-grid effects into a Europe-wide health impact assessment for PM/Ozone?

7. The JRC has coordinated: A model inter-comparison exercise for urban-regional scale dispersion models focusing on 8 European cities to identify:  the systematic differences (delta’s) between rural and urban background AQ (“Scale”),  how these delta’s depend on emissions (“Emissions”),  how these delta’s vary across cities (“Cities”),  how these delta’s vary across models (“Models”)  how these delta’s vary for PM and O3 (“Pollutants”).

8. Driving force : WHO Review of health impact from air pollution Damage from long-term exposure to PM2.5/PM10 Not yet possible to distinguish potency of individual PM components No threshold value for health impact can be identified The levels of long-term mean concentrations are more important than (episodic) peak concentrations  CityDelta - Indicator: Annual mean PM2.5/PM10 (W/S) New evidence for mortality effects from ozone No firm evidence for no-effect level; but larger uncertainties for effects at low concentrations Thus also low ozone days are relevan  CityDelta - Indicator: SOMO35 (*) (*) Sum of exceedances over 35 ppb of the maximum of the daily 8-hour mean O3 concentrations, calculated over the entire year

9. Model# of levelsLevelDomainResolution CALGRID11 10m CITYDELTA 5-10 CAMx1110mCITYDELTA5-10 CHIMERE local6: surface-700hPaSL (50m)CITYDELTA5 CHIMERE regional6: surface-700hPaSL (50m)Europe50 EMEP Unified Model20: surface-100hPa1mEMEP50 EMEP-v.120: surface-100hPa45mEMEP50 EPISODE6: 25-2500m2mCITYDELTA10 EUROS425mCITYDELTA10-50 LOTOS local3: 0-3500mMLCITYDELTA5-10 LOTOS régional3: 0-3500mMLEurope50 MOCAGE47: surface-5hPa1st level (0-50m)Paris, Milano10-50 MUSCAT22: 0-4400m1st level (0-33m)CITYDELTA10 MUSE510mCITYDELTA10 OFIS2MLCITYDELTA5 REM3 local4: 0-3000mSLCITYDELTA5 REM3 regional4: 0-3000mSLEurope50 STEM-FCM1110mCITYDELTA5 TRANSCHIM1050mCITYDELTA5-10 15 Modelling teams: 7 regional-scale 11 urban-scale

10. 8 Cities: London Paris Prague Berlin Copenhagen Katowice Milan Marseille

11. 8 Emission Scenarios 0 --- 1999 1 --- 2010 CLE: Current Legislation 2 --- 2010 NOx MFR: Maximum Feasible Reduction 3 --- 2010 NOx (CLE+MFR)/2 4 --- 2010 VOC MFR 5 --- 2010 NOX and VOC MFR 6 --- 2010 PMcoarse MFR 7 --- 2010 PM2.5 MFR NOx CLE-1999MFR-1999 Prague-34%-62% Milan-36%-53% Paris-42%-65% Berlin-38%-50% Meteo: 1999 provided by Meteo-France (Aladin 10 km) or calculated. Boundary conditions: provided by EMEP or calculated. Long term simulations: full year for PM, 6 months for O3 Outputs delivered with resolution of 5-10 or 50 km

12. Emission Inventories: Local vs Regional NOx, CO, SOx estimates seems quite robust PM estimates show 40-50% differences.  CityDelta has also contributed to a considerable revision of the regional-scale (EMEP) emission data

13. Model Validation: - The “Taylor” plots - The “Ensemble” model CRMSE Corr. K.E.Taylor, 2001, JGR, 106, 7183-7192 O3 Bias RMSE

14. O3 Summer Mean Fine scale Ensemble Large scale Ensemble BER MIL LON KAT PRA PAR PM10 Winter BER MIL PRA PAR “Delta” Interpretation (City-Model)

15. “Delta” Interpretation (Emission-Scale) NOx Reduction VOC Reduction CLE-2000 SOMO35: (CityDomain) Fine scale Ensemble Large scale Ensemble

16. CLE-2000 NOx Reduction VOC Reduction “Delta” Interpretation (Emission-Scale) Fine scale Ensemble Large scale Ensemble PM10: (CityDomain)

17. Discussion A first approach for addressing urban air quality for Europe- wide health impact assessment has been developed and implemented – based on observations and City-Delta results Urgent need for validation of the FR with monitoring data, hampered by lack of PM2.5 sites. Presently, grid average wind speeds used. No consideration of topography. City-specific wind speeds should improve. Which emission/population density is representative of a city (how to draw city borders)? This determines directly the size of the urban increment.

18. Publications: Overview: Cuvelier,Thunis, Vautard, et al. Validation: Vautard, Thunis, Cuvelier, Builtjes, et al. Delta’s: Thunis, Rouïl, Cuvelier, et al. Emissions: Maffeis, et al. All: Atmos. Env.

19. EuroDelta

20. Project to evaluate uncertainty in SRRs used in AQ policy – 5 regional models: EMEP, MATCH, REM3, CHIMERE, LOTOS – 28 emission scenarios (2000, 2010, 2020) with area specific reductions (FR, GE, IT, NorthSea, MedSea) – 60 sectoral emission reductions (2010, 2020) in FR, SP, GE, UK, and MedSea –Zoom studies for Marseille, and Athens (50km => 10 km) – Use of the ENSEMBLE approach – Objectives : - Source-receptor variability - Spatial variability - Meteorological variability - Confidence limits for policy purposes

21. BaseCase: Emissions 2000, Meteo 1999 –Differences between model results of about 10 ppb in the rural areas and over the countries ; largest differences in winter –Model variabiliy of 10 ppb represents about 40% of the ensemble value –Models are closer in highly populated areas and more different in rural areas and over the seas Mean Ozone max ensemble min Countries

22. Main Publications : Validation: in Atm. Env. Delta’s: Report EuroDelta I (to appear) Report EuroDelta II (Sep 2008)

23. http://aqm.jrc.it/eurodelta Report + DVD available with Results http://aqm.jrc.it/citydelta DVD available with Results

24. Need for Harmonisation Need for Evaluation of the uncertainty in AQ modelling used for AQ policy. Do AQ models produce consistent responses to emission-reduction policy scenarios; What are the ranges of these responses ?  What is the robustness of AQ modelling results in the policy context?

25. Harmonisation as we see it (and as Fairmode may see it) Analysis of individual Modelling results (Validation against Obs) Intercomparison of Modelling results Ensemble approach (mean/median of the Models) Analysis of the variability around the Ensemble Validation of the Ensemble model Spin-off: improvement and further development of models The Ensemble model is a useful tool in the frame of the Harmonisation of the individual model results. It is often more robust and looks therefore more appropriate for policy purpose 

26. The JRC Evaluation-Visualisation Tool

27. OBSERVATIONS (surface, profiles, plane, lidar… MODEL 1MODEL 2MODEL 3, … 4-D Fields: Meteo, Chemistry, Aerosols Pre-processing (Window Interface) ASCII-BIN  CDF MONIT Observation Overview SHOWALL 4D model data analysis XZ, YZ, SZ, XY Sections Comparison with obs. VALID Statistical Module Multi-models (obs.vs. models)

28. MONIT Station Overview Station raw data

29. SHOWALL

30. VALID